Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Serum Cholesterol Response to Changes in the Diet. IV. Particular Saturated Fatty Acids in the Diet ByANGEL KEYS, JOSEPHT. ANDERSONAND FHANCIXO GRANDE For many dietary changes satisfactory prediction of the average change in the serum cholesterol level of man in mg./ 100 ml., is given by a Chol. = 1.35 (2nS - AP) + 1.5AZ where S and P are percentages of total calories provided by glycerides of saturated and polyunsaturated fatty acids in the diet and Z” = mg. of dietary cholesterol/ 1000 Cal. This formula fails, however, when the dietary change involves large amounts of cocoa butter and discrepancies also appear with beef tallow or hydrogenated coconut oil diets. Controlled dietary experiments at the Uni- versity of Minnesota and at 2 other centers, provide 63 sets of comparisons of serum cholesterol averages for groups of men on each of ,2 chemically characterized diets. Least-squares analysis indicates that stearic acid, as well as saturated fatty acids containing fewer than 12 carbon atoms, have little or no effect on serum cholesterol in man. The equation, @, Chol. = 1.2(2aS’ - AP) + 1.5~Z, yields good correlation (r = 0.93) with the observed values in these 63 sets of data. This formulation also resolves heretofore puzzling discrepancies in the literature. F OR ALL ORDINARY DIETS, as well as for most experimental diets. reasonably satisfactory estimates of the average serum cholesterol response to changes of fats in the human diet can be made from data on the percentage of total calories provided from saturated, S, and polvllnsaturated, P, fatty acids in the diets concerned ( cf. the previous parts of the present communication) .lvt’ The formula previously used for this purpose, A Chol. = 2.7nS - 1.3AP, or 1.35(3/Q - AP) implies that, in this regard, all saturated fatty acids are equivalent, but this formulation was designed only as an approximation to apply to most practical dietary situations without implications as to particular fatty acids. In other words, the coefficient +r1.7 applies to mixtures of the saturated fatty acids as they were represented in the considerable variety of diets used in the derivation of the formula. The limitations of 1.35 (2~s AP) are indicated by the experimental diets studied. These included common U.S. ‘luxus,” medium- and low-fat diets of ordinary foods with and without inclusion, in amounts up to about 35 per cent of total calories, of the following experimental fats: corn, soybean, coconut and olive oils, sunflower seed, rapeseed, safflower, cottonseed, “sardine” (pilchard) and menhaden fish oils, and butterfat. From other data it appears that the formulation should also apply to sesame, peanut and mustard seed oils. With the exception of diets containing unusually large amounts of coconut From Laboratory of Physiological Hygiene. Minnesota. Aided by research grants frtml the Cl. S. Public M. HE-04401 to F. G. and J. T. A.) Received for publication Feb. 19, 1965. 776 Crniversity Ifedth of Socicc Minne.sotci, Rlinneapoli,s, i no. HE-049Y7 to A. K., 777 SERUMCHOLESTEROLRESPONSETOCHANGESINDIET oil and butterfat, in almost all diets palmitic acid dominates the saturated fatty acid group, and in effect, the term “S” mainly refers to this fatty acid. In coconut oil, lauric acid is dominant but fatty acids with fewer than 12 carbon atoms make up about 16 per cent of the total saturates. In butterfat, such short- and medium-chain length fatty acids make up around 10 per cent of the saturated fatty acids. There is good reason to believe that, in the diet, fatty acids with fewer than 12 carbons have much less effect on serum cholesterol than the longer chain saturated fatty acids. We were unable to find any effect of butyric acid in the diet.” Hashim et al.” found virtually no serum cholesterol effect of a mixture of fatty acids with fewer than 12 carbons in the chain, even when fed in large amounts to human subjects. Beveridge et al.” provided confirmation. The fatty acids with fewer than 12 carbon atoms in the chain are more polar and less hydrophobic than the other fatty acids and they appear to be metabolized quite differently, being absorbed via the intestinal capillaries and the hepatic portal system rather than via the lymphatics.6-9 It is reasonable, therefore, to suggest exclusion of such short-chain fatty acids from the S term in estimations. This would have little consequence for calculations except in the case of diets specially devised to emphasize the short-chain fatty acids. In the case of butter, its cholesterol-raising effect (independent of the cholesterol in the butter), may be overestimated by around 10 per cent if all saturated fatty acids are included in the term S. Similarly, in the case of coconut oil, inclusion of all saturated fatty acids in the S term may lead to a small overestimation of cholesterol effect. At the other end of the scale of chain length among saturated fatty acids, only stearic acid is ever present in the diet in considerable amount. Tristearin is very poorly digestible and it would be expected, therefore, that the effect of tri-stearin on cholesterol metabolism might well differ from that of the more usual fatty acid glycerides. But tri-stearin is present in most natural food fats only in trivial amounts and it would seem safe to neglect consideration of it. Triglycerides containing one or two stearic acid molecules are digestible and are less uncommon in food fats but until recently we saw no reason to distinguish between stearic and palmitic acid, for example, in the diet. However, the pursuit of some discrepancies between observation and predictions from 1.35(2aS - nP) led to an unexpected picture. EXPERIMENTS WITH COCOA BUI-IZR In connection with studies on the effect of dietary cholestero1 on the serum cholesterol level, it was desired to make comparisons in diets containing highly saturated fatty acids. For this purpose, cocoa butter was a convenient choice to use in the fat mixture for the saturated fat diet. In the outcome, the effect on the blood of the dietary cholesterol was identical in various types of diets (vide supra, Part II), but the serum cholesterol levels did not correspond to expectations when the 2 diets were compared at the same level of cholesterol intake. For example, in experiment AF, the formula 1.35(2~8 AP) predicted an average difference of 65 mg./lOO ml., but the difference observed 778 KEYS. ANDERSON AND GRANDE was only 33; this is far too big an average discrepancy for fully controlled experiments on 22 men subsisting on each of the 2 diets in a switch-back design. Similar discrepancies appeared in the AM series in which cocoa butter was also used in one of the diets. Experiments AF and AM were completely comparable, except for the use of cocoa butter in the diet, with the long series of previous studies which were also conducted in the Metabolic Unit of the Hastings State Hospital with physically healthy, schizophrenic men. Transfer of the Metabolic Unit to the Faribault State School and Hospital provided opportunity to organize new experiments with men of a very different type; these subjects were also physically healthy but were mentally defective rather than psychotic. The FB series of experiments at Faribault also indicated a similar peculiarity in serum cholesterol response. While these puzzling findings were under scrutiny, Connor et al.“’ published their data from similar experiments with dietary cholesterol and their results were almost identical with those we had obtained, both in regard to the effect of dietary cholesterol and in the unexpectedly low serum cholesterol levels when the diet included substantial amounts of cocoa butter. While preparations were under way for further experiments designed to provide critical tests of hypotheses to explain the peculiar results when the diet contains cocoa butter, the paper by Erickson et al.” appeared. They too used substantial amounts of cocoa butter in the diet and their data are in agreement with the findings here and at Iowa City. The outstanding peculiarity of cocoa butter, as indicated by its detailed analysis, is its extraordinarily high content of stearic acid; stearic acid makes up an average of 3-5 to 36 per cent of the total fatty acid in cocoa butter. For comparison, the percentage of stearic acid in other fats averages as follows: mutton tallow, 30; beef tallow, 20; milk fat, 10; lard, 13; and most ordinary vegetable oils 2 to 4. In usual American diets stearic acid seldom accounts for more than 3 per cent of total calories. Several explanations may be sug,gested for the serum cholesterol discrepancies observed when cocoa butter was present in the diet. Conceivably, the active ingredient could be something other than stearic acid, and this possibility deserves investigation. But it is useful to consider the possibilities concerning stearic acid. (I) Perhaps stearic acid has no cholesterol-promoting effect and simply should be omitted in computing S in the formula, 1.35(2OS - AP). If stearic acid glyceride is denoted by S” and S = S’ + S”, the formulation could be suggested: AY = 1.35( 2AS’ - ar , (IV, 1) in which Y = serum cholesterol (mg./lOO ml.), and S’ = S - S”, i.e., the S’ = saturated fatty acids excluding stearic acid. (2) Stearic acid might have a direct cholesterol-lowering effect, something like that of polyunsaturated fatty acids, in which case the formulation could be: (Iv, 2) AY = 1.35(2nS’ - AP) - ea S”. where e is a constant. SERUM (3) CHOLESTEROL Alternatively, RESPONSE TO CHANGES 779 IN DIET it could be suggested that stearic acid interferes the effects of the other fatty acids on the serum cholesterol. expressed by the equation: (Iv,31 AY = 1.35(2AS’ with This hypothesis is - AP) (1 - eAS”). Conceivable variants on this latter idea would be that stearic acid interferes with the action of S’ or that it potentiates the action of P. Designs are under consideration for experiments to test these different hypotheses critically. In the meantime, there are available many data suitable for statistical analysis to indicate the possibilities, including practical estimation of the results of dietary changes involving stearic acid. Substantial differences in stearic acid content were not involved in the dietary comparisons we have analyzed previously.ls* However, all natural diets contain glycerides in which stearic acid has at least a small representation. Accordingly, it seemed desirable to reanalyze all of the data from controlled experiments in which reliable dietary details are available. The experiments summarized in table 1, conform to these requirements. In most cases, the fatty acid composition of the diets was determined by gasliquid chromatography of the total lipid extract from homogenates of the entire diet as eaten. In the other cases, a constant low-fat basic diet was used to which was added experimental fats of known composition, the fatty acid composition of the basic diet being estimated from tables of average food composition. Accordingly, though in some of the experiments there may be questions about the precise amount of stearic acid in the total diet, within any one set of experiments the differences between diets in stearic acid content are accurately known. Apart from the foregoing reason for making the analysis in terms of differences between diets within the separate sets of experiments, the men and other conditions were not the same in the various sets of experiments, though these variables were constant within each set. The subjects studied by Connor et al.1° included several controlled diabetics as well as healthy volunteers; the subjects of Erickson et al.li were prisoners in a penitentiary. And, as noted earlier, the Minnesota subjects were schizophrenic men in one state hospital and mentally defective men in another. For table 1, we did not include experiments in which fish oils provided most of the fat calories because of lack of adequately detailed data on chemical composition directly obtained from the diets as eaten. However, in all, table 1, provides data allowing 63 comparisons between the averages of groups of men where the only variable was the dietary situation indicated in the table, It seemed desirable, therefore, to make no assumptions from the previously obtained coefficients, +2.7AS and -1.3AP, and to make a new, independent, multiple regression analysis. The following models were analyzed: AY = bA.9 + dAP, (IV, la) (IV, 2a) (IV, 3a) AY = bAS’ + dAP f AY = (bAS’ + dAP) (1 - eA S”, e AS”). 780 KEYS, ANDERSON AND GRANDE Table 1 Line Reference N F P Z’ Serum 1 Minn. HWX-LF 13 11.4 x :3 I.4 1.5 94 2 Minn. HWX-CT 1 :i 37.0 Y.0 1 .Y 14.8 94 1 B(I 3 Minn. HWX-CO 13 37.4 5.6 1.!1 16.3 !I4 160 1s 4 Minn. HYZ-LF 12 11.x 8.4 1.4 1.6 94 1X1 5 Minn. HYZ-HYD 12 37.6 20.0” 3.8 0.9 Y4 224 G Minn. HYZ-CT 12 37.5 9.2 1 .Y 14.s !I4 IiX 7 Minn. JWX-LF 14 11.3 :3.i 1.4 94 1X5 8 Minn. JWX-OL 14 37.4 6.0 l.Y !!4 188 9 Minn. JWX-SU 14 37.0 Ii.6 1.9 !I4 17u 10 Minn. JWX-CO 14 87.6 5 7 I.9 Y4 161 11 Minn. JWX-RU 14 36.X 14.4* :3.7 12 Minn. JYZ-CT 12 39.2 1G.l 1 .!t 13 Minn. JYZ-01, 12 39.3 Ii.2 1.9 14 Minn. JYZ-CO 12 39.0 5.9 I.!1 15 Minn. JYZ-BU 12 3X.X 1.5.6* :i .!I 16 Minn. NWY-SA 12 :33.4 6.3 17 Minn. NWY-CO 12 38.9 7.6 18 Minn. NXZ-SA 12 38.3 Ii.3 19 Minn. NXZ-CO 12 :3x.4 i.4 20 Minn. P-LF 22 21 Minn. P-OL 22 R6.3 6.7 22 Minn. P-SM 22 1X.3 6.5 ‘i !j.F, :3 214 224 1 .8 !,4 ltil 1 .!I 94 15Y 1 .!) !I4 I .57 1 .9 !I4 162 I).!1 94 15!l 1.6 !I4 171 1.9 ‘14 lil 23 Minn. AM-LCS 22 8.1 x.1 0.7 17 205 24 Minn. AM-LCF 22 3x.9 21.x 6.1 li 239 25 Minx AM-HCS 22 X.1 3.1 0.x Sfifi 21Y 26 Minn. AM-HCF 22 40.1 21.1) 7.1 545 259 27 Minn. AF-LCF 22 40.6 24.2 5.4 1X 22x 28 Minn. AF-LCE 22 40.9 10.0 1.7 1X lY4 29 Minn. AF-HCF 22 39.8 23.7 5.7 525 257 30 Minn. AF-HCE 22 39.6 9.x I.7 523 224 31 Minn. FU-SPBO 28 32.9 8.9 1.8 !J4 211 32 Minn. F&OS28 28 15.6 X.5 I,.9 !14 206 33 Minn. FB-CHO 2x 7.0 2.6 11.x 94 209 34 Minn. FB-OS80 2x 33.Y 3.i 1 .3 94 35 Minn. FB-BU 28 33.0 13.7* 4.1) 210 36 Conn. I 5 40.0 9.1 6.9 37 Conn. II 5 B9 7 .x 7.6 1 XC) 38 corm. III 5 40.0 6.0 1.4 174 39 Conn. IV 5 39.9 ii.!1 1 .% 202 !) 40 Eric. A 21 40.4 5.5 2.7 41 Eric. A-t 22 40.4 5.5 2.7 42 Erie. B 21 40.5 5.9 2.7 43 Eric. B+ 19 40.5 5.9 2.7 44 Eric. C 21 40.7 i .:i 6.2 45 Eric. D 20 40.8 7.:3 6 :2 46 Erie. E 20 40.4 *Saturated Mean fatty values chol~terol/lOOO glycerides unsaturated of acids for Cal. stearic fatty with serum (=Z?), acid acids fewer (mg. (=P). 12 carbon cholesterol/100 and percentage of (=S”), N than 10.2 = of saturated number of atoms ml.) and are not calories fatty acids Conn. 0 :30ti 0 XUli 0 0 0 provided other = 193 “17 18X 215 190 188 196 -__- included composition total men. ““2 12.7 for ZOti 244 than Connor of by the total stearic et al.“’ Eric. diet as fat (z-F). (=S’j. and - eaten: Erickson mg. and of by polyet al.” SERUM CHOLESTEROL Expansion (IV, 3w RESPONSE TO CHANGES of equation AY 781 IN DIET (IV, 3a) yields an equation with cross-products = baS’ + dnP- eba SPAS” - ednP AS”, and no single unique value for the coefficient e can be guaranteed model. In the actual analysis the equation is (IV, 34 AY = bAS + dAP - f asasfr - in a linear gAPAS”. Many of the experiments summarized in table 1, involved no differences in dietary cholesterol, Z2, but in others account must be taken of Z2 if valid comparisons are to be made. For this purpose we have used the estimate of dietary cholesterol effect previously obtained, i.e., a Chol. = 1.5~2, when all other conditions are constant. Hence in those comparisons in which AZ P 0, the term 1.5~2 was added to each of the equations above, Z2 being mg. of cholesterol/1000 Cal. The coefficients for saturated (less stearate) and for polyunsaturated fatty acid are + and -, respectively, in every solution and the ratio of these 2 coefficients varies only from 1.70 to 2.34, the average being 1.99. This means that the controlling variable must be very close to 2s’ - P. The least-squares solution for the simplest model, equation (IV, la), may be written as n Chol. = 1.1 (2.05~s’ - AP). A close approximation, which we propose for general use, is the still simpler form (IV, lb), n Chol. = 1.2(2aS’ - aP). Where AZ is not zero, the value +1.5~2 must be added to any of the above-noted equations, of course. The improvement resulting from distinguishing between saturates less stearate, S’, and simply saturates, S, may be seen when the comparisons from table 1, are used to calculate the correlation coefficient between ‘2~s - AP and the observed a Chol.: the result is r = 0.91 for the Minnesota data and r = 0.88 for all 63 sets of comparisons. These values are in contrast with the corresponding values r = 0.96 and r = 0.93 when the calculation is made with 2AS’ - P. In all of these calculations the observed serum value was corrected by 1.5AZ when AZ was not zero. For the 40 sets of comparisons where AZ = 0, i.e., no difference in dietary cholesterol was involved, the coefficients of correlation of A Chol. with 2s -.- P and from 2s’ - P are r = 0.91 and r = 0.96, respectively. These improvements in the coefficient of correlation are more consequential than might appear at first glance. With r = 0.96, only 8 per cent of the variance remains unexplained but with r = 0.91, the unexplained variance is 17 per cent, i.e., more than twice as great. And when r = 0.88, variance explained is only 77 per cent of the total. More important from the standpoint of theory is the fact that results in some experiments, e.g., with cocoa butter and ethyl stearate, are predictable from 2s’ - P but otherwise would be wholly discrepant. Figure 1 compares the observed values for n Chol. with those predicted from equation (IV, Ic), i.e., n Chol. = 1.2(2AS’ - AP) + 1.5~2. It is instructive to compute the coefficient of correlation between the 782 KEYS. ASDERSOK ASD GRhNDk: C 70 A CHOL.=%2(2AS'-AP)+1.5A2 0 MINNESOTA . CONNOR et al Y ERICKSON et cl B AHRENS et al. Fig. L-Correlation between a cholesterol observed and the values predicted from 1.2(2aS’ - AP) + 1.5~2. Each point is the average for a group of meli maintained on each of 2 diets. Open circles ;ue Minnesota d;it;l, solid circles from Connor et al.,“’ half solid from Erickson et al.,” cross from 2 patients fed COCO;L butter vs. butter.lz observed serum the iodine value r = where -0.23 observed these cholesterol for less than except 10 per cent where in stearic -0.28 where needed, indicate between and significant i.e. AS” where and is much Q is result correcting calculations. correlations. If correction function AZ = 0. But the correlation acid is involved, AFQ, for all comparisons, of the variance. value function from total fats. The as in the previous statistically the iodine the of calories AZ = 0 and r = values by l.SnZ coefficients correlation differences and F is percentage they While account for AZ is omitted, A ChoI. better is the the is not significant when no difference = 0. DISCUSSION The experimental analysis with 7 groups data reservations because of 4 to 6 men each so that the cholesterol of subjects. of Erickson values In other words, et al.,” of the were included experimental in the design. present There were and 7 diets were fed but in only 4 experiments on each diet represented in comparing average different combinations values on any different diets about half of the men were the same on the two diets but the other half of the men were different on those two diets and the “treatment effect” is combined 783 SERUM CHOLESTEROL RESPONSE TO CHANGES IN DIET Table 2 A Chol. = (IV, la) SOUrCe Diet. Chol. Minn. All N 30 2.063 -1.159 1.78 0.97 2.281 -1.269 1.80 0.96 6.66 AZ=0 40 1.974 -1.008 1.9fi 0.94 6.31 63 -__--_...~_ 2.249 -1.09x 2.05 0.93 7.5x All + azAP + n-.&W 30 2.626 -1.199 -2.331 2.19 0.97 4.84 All 45 3.048 -1.305 -2.960 2.34 0.96 6.:39 AZ=0 40 2.442 -1.202 -1.583 2.03 0.96 5.05 All 63 2.762 -1.324 -1.831 2.09 ..-__-- 0.95 6.35 -~ axAS’ + w:AP + a:rAS’AS” + a,APaS* N a, r 30 2.543 -1.299 -___ 0.056 al/a? AZ=0 0.370 1.96 0.93 3.97 All 45 2.521 -1.409 0.074 0.270 1.79 0.98 5.22 -0.115 1.70 0.95 5.x1 -0.061 2.18 0.94 8? AZ=0 40 1.779 -1.044 All 63 2.251 -1.033 Least-squares of the _^__ 5.01 AZ=0 A Chol. = Minn. S.E.E. 1‘ 45 3c) between a/a? A Chol. = a&S’ All (IV, a? Ali 2at Minn. TV a:hP AZ=0 All (IV, a, a&S solutions experiments regression: for within r = equations each coefficient IV, la, of the of correlation 12 sets a1 84 -0.067 0.065 IV, 2a, listed between and in table IV. 3c using 1. S.E.E. observed and = the data standard predicted on S.E.E. all en-m 7.10 differences of estimate values. with interindividual differences. Further, the cholesterol values are reported without indicating confidence limits and had been “adjusted” (for time trends?). Actual observed means and standard deviations were not reported for any of the 7 groups on any of the 7 diets or for the control period. Proper detailed calculations cannot be from the published data. From the indicated significance of the adjusted means for groupings, it can be inferred that a mean difference of 20 mg. cholesterol/100 ml. serum would be significant at p = 0.05 or less in that material. In spite of such limitations, comparisons were made between observed differences and those predicted from 1.2(2aS’ - AP) + 1.5AZ. Twelve sets of differences are available from the material and for these the coefficient of correlation between observed and predicted differences is r = 0.87. Treating the observed and predicted n values as 2 estimates of the same thing, the standard error of measurement is S.E.M. = 57.1, a value that must be similar to the standard errors between the observed means. It is concluded that the data of Erickson et al.” are consistent with the hypothesis that the average serum cholesterol response to a change in lipids in the diet is predictable from 1.2(2AS’ - AP) + l.sAZ, where S’ does not include stearic acid. Besides the experiments covered in tables 1 and 2, scattered data in the literature can be examined with the help of table 3, which gives the approsimate average composition of important food fats and oils. While different samples of such fats vary considerably, these values provide the basis for rough calculations for the experiments mentioned below, as well as for other purposes. Ahrens et all2 fed patients No. 26 and No. 37 on 40 per cent fat calories, using both ordinary butter (BU) and cocoa butter (CB). Though the BU diet 784 EEYS, ANDERSON AND GRANDE Table 3 l*‘at l)iet (‘hc>l. HO Beef tallow X30 Bntterfat Cocoa bu ttc1 Coconut 0 oil (I IO0 Beef tallow Coconut oil, hydrog. Corn oil Cottonseed Lard 0 0 oil 0 100 Mustard seed oil Mutton tallow 0 90 Palm oil Peanut butter* I) Peanut oil Olive oil I) Rapes4 oil Safflower oil Sesame oil 0 Soybean (1 ‘Lightly 0 0 II (I oil hydrogenated. Approximate average characteristics of ctlihk fats ant1 oils. Q = iodine value; (71~01. = mg. choIesterol/lOO g. s’, s” ;mtI p are percentage5 of total tat rc~pr~smtrd Ity glyceridrs ot. respectively. polynnsaturated saturated fatty fatty acids. fat had a slightly more acid\ higher iodine polyunsaturated mg./lOO ml. higher in table expected than fatty 11. to 16 carbon> value acid, ill the and provided the average chain. straric less saturated ncitl. and ;I little scArurn cholesterol on the BU than on the CB diet. The dat;i ;~ntI was are summarized et al..‘” it would 11 4. Judging from iodine values, that the serum on the stearic with CB diet. acid predicts And the BU - by Ahrens level would calculation be slightly ignoring lower dietary from the equations agreement with the observed developed difference. the predictions for BU - in the From present equation paper IV, CR are a Chol. and in table gives lb, bc on the BU cholesterol CB = -:3 mg./lOO ml. But, as indicated prediction (for all 63 sets), emphasized cholesterol 1, good and IV, 3c, = 40.8 and n Chol. = 33.7, respectively. Ahrens high stearic observed, (28 No. 29 and No. 35, corn oil and beef tallo\\ et al.” also fed 2 patients, at 40 per cent fat calories acid and these experiments content The serum was n minus corn oil diet, The calculation n Chol. diet = 79 which becomes of cl~olestero1/1OO0 Cal. in the tallow The equations developed in this 89 when paper patients but to a much smaller (IV, a Chol. = 54; a Chol. (IV, 3c), also = nP) is made = i12 would for 40 mg. = 9.5). overpredict degree: - of thr difference 27, S.E. 1.35(2AS allowance X 6.3 because cholesterol Chol. from,” diet (1.5 tallow-fed la), are interesting tallow. -C 10.9 and 26 -C 13.2). predict tallow of beef (IV, a lc), ChoI. for n Chol. = 44. But it must be noted = the 58; that SERUM CHOLESTEROL RESPONSE TO CHANGES Table Diet I.V. --__ BU CB 39.5 36.6 IV, 3c S” 40 40 19.5 10.0 14.0 from equations: n Chol. = 1.2( 19.0 - 0.4) Chol. = 2X5(9.5) -0.06( Means for 2 patients studied (CB) diets. Symbols as in table Cal. of diet. -10.4) 785 4 S’ 3.6 9.5 n DIET F _____ .A Predicted IV, lb IN -10.4 + 1.5( 12.3) 1.03(0.4) (0.4) by Ahrens 1. I.V. = z P z Chol. 1.6 1.2 12.3 0 245 204 0.4 12.3 41 40.8 + 0.06( -10.4)(9.5) + 1.5( 12.3) = 33.7 et al.12 on butterfat iodine value. Z 2 = (BU) 152 mg. and cocoa butter cholesterol/l000 these 2 patients were relatively hypocholesterolemic and adjustment must be made accordingly .la If, on the average, these patients had about 72 per cent of the reference serum cholesterol value on a reference diet, i.e., 7 = 0.72, the prediction from equation (IV, lc) would be a Chol. = 28, as will be seen from Keys et al. I8 In any case, allowing for stearic acid in the diet much improves the prediction. Table 3, suggests that it would be interesting to examine experimenta data for diets containing a large amount of hydrogenated coconut oil such as used by Malmros and Wigand. 1-LExact numerical analysis is impossible but it is noted that in their experiments the serum cholesterol levels did not rise significantly when their “free” diet was replaced by a diet in which hydrogenated coconut oil was the only fat and provided 40 per cent of the calories. A reasonable estimate for an average “free” diet in Malmo, Sweden, is S = 17, S’ = 13, S” = 3, P = 4 and 250 to 300 mg. cholesterol/1OOO Cal., i.e. Z = 17. The hydrogenated coconut oil diet provided 60 per cent of the calories from “bread, cereals, vegetables, potatoes, rice, fruit and sugar” (op. cit.) so that the total diet must have been about S = 37, S’ = 28, S” = 4, P = 3 and Z = 0. Allowing for AZ, equation (IV, lc), predicts a Chol. = 11 mg./lOO ml., an expected difference that would be difficult to demonstrate with small groups of subjects. Finally, we recall Horlick’sl” experiment in which ethyl stearate added to the diet failed to produce a rise in serum cholesterol as expected from the increase in saturation of the dietary fat. This long-puzzling oddity is entirely in keeping with the current analysis. All available data are consistent with the theory that stearic acid in the diet is not cholesterol-promoting. There is little reason to suggest that some mysterious substance in cocoa butter is responsible for the uniformally surprising result when that fat is a prominent part of the diet. Except for the dominance of stearic acid, there is no obvious peculiarity about the fatty acids in cocoa butter; it contains practically no short-chain fatty acids, and the average composition is palmitic = 25, stearic = 35, oleic = 37, linoleic = 2 per cent according to Eckey;16 samples analyzed by us are in agreement except that we find more linoleic acid, as high as 8 per cent. The unsaponifiable frac- tion of cocoa butter stance were responsible, action. But the best fact that other acid-beef to the per cent argument against cocoa on serum of variation of calories coconut suggest an extremely agent which in cocoa oil-seem sub- powerful butter are relatively of the data in table in the diets from total fats. IIre have on students have a special butter from the analysis range must is the rich in stearic to exhibit a similar cholesterol. high fat diets but unpublished ments amount and hydrogenated in effect Generalizations fined only from 0.2 to 1 per cent. If some unknown a tiny fats besides tallow peculiarity ranges data obtained that 1, are properly represented, little i.e., from experience with cxtremel\ by Dr. H. L. Taylor the relationships shown con- 7 to -11 here in experi- extend also to who were fed formula diets in diets as high as 53 per cent in fat calories. Ahrens which et a1.l” reported corn oil provided data from 4 patients both was extraordinary in that all diets, including variations patients, extra his severe corn oil, the averages respectively, current being fatty analysis suggests is much dietary palmitic diets, acid it may no significant cholesterol No. 13 same response corn oil. Perhaps there on thrct to the is an upper that the cholesterol-promoting acids. Since the limit acid effect. human in most was 165 and 172 mg./lOO ml., on 40 and on 70 from rated fatty acids in natural acids Patient from 10 to 70 per cent corn oil. The other calories for the polyunsaturated fat calories. hvl~er~holesterolemia Nos. 31, 32, and 34, showed per cent, The 40 and 70 per cent diets is due to lauric, more abundant be primarily fat to the serum cholesterol responsible than effect myristic lauric for the of satll- and pahnitic and myristic contribution of level in man. ACKNOWLEDGMENTS Acknowledgments of help in connection I of this communicatiorl. Joan The statistical with thcb c,xpc+mc,ntal stlldies work was aided by hlr. R Willis arc given in Part Parlin ant1 bliss LaRiviere. REFERENCES 1. Keys, A.. Anderson. J. T.. and Cmnde. F.: Prediction of serum cholesterol responses of man to changes in fats in the diet. Lancet ii:959. 1957. and -: Serum cholesterol in 2. -, -, man: diet fat and intrinsic responsivrness. Circlllation 19:201, 1959. 3. -: Chemistry of Lipides as Related to Atherosclerosis. Ed. by I. H. Page. Springfield. Ill.. Charles C Thomas. 1958. p. 252. -1. Hashim. S. A.. Arteaga. A.. and van Italie. T. B.: Effect of a saturated medium-chain triglyceride on serum lipids in man. Lancet 1: 1105. 1960. 5. Brveridge. J. M. R.. Connell, W. F.. Haust. H. L.. and Mayer. C. A.: Dietary cholesterol and plasma cho- 6. 7. 8. 9. kstcrol levels it1 1)un. Canad. J, Birothem. Physiol. 37: 575. 1959. Blomstrancl. R.: Transport form of dccanoic acid-1-1X in the lymph during intestinal absorption in the rat. Acta Physiol. ~cand. 34: 67. 1955. Bloom B.. Chaikoff. I. I,.. and Reinhardt. \V. 11.: Intestinal lymph as pathwa! for transport of aborbecl fatty acids of dift’rrent chain lengths. Amer. J. Physiol. 166:451, 19.51. Fernandez. J.. van de Kamer, J. H.. and Weijers. II. .4.: The absorption of fats stlldied in a child with chylothorax. J, Clin. Invest. 34:1026. 1955. (:r:mtle. F.: Dog serunl lipid responses to dietary fats differing in the chain length of the saturated fatty acids. J. SERUM CHOLESTEROL RESPONSE TO CHANCES Nutr. 76255, 1962. 10. Connor. W. H.. Stone. D. B.. and Hodges. R. E.: The interrelated effects of dietary cholesterol and fat upon human serum lipid levels. J. Clin. Invest. 43:1691, 1964. 11. Erickson. B. A.. Coots, R. H., Mattson F. H.. and Kligman. A. M.: The effect of partial hydrogenation of dietary fats, of the ratio of polyunsaturated to saturated fatty acids, and of dietary cholesterol upon plasma lipids in man. J. Clin. Invest. 43:2017. 1964. 12. Ahrens. E. H.: Jr.. Hirsch. J., Insull, W.. Jr.. Tsaltas, T. T.. Blomstrand. R.. ancl Peterson, M. L.: The influence of dietary fats on serum-lipid levels in IN DIET 787 man. Lancet 1:943, 1957. 13. Keys, A., Anderson, J. T.. and Grande, Serum cholesterol response to F.: changes in the diet. III. Differences among individuals. Metabolism 14: 766, 1965 14. Malmros, A.. and Wigand, fect on serum-cholesterol taining different fats. G.: The efof diets conLancet 2:l. 1957. 15. Horlick, L.. and Craig, B. M.: Effect of long-chain polyunsaturated and saturated fatty acids on thr serum-lipids of man. Lancet 2:566, 1957. 16. Eckey. E. W.: Vegetable Fats and Oils New York, Amer. Chem. Sot. Monograph Ser. Reinhold. 1954. And Keys, Ph.D., Professor, School of Public Health, Director, Laborutory of Physiological Hygiene, University of Minnesota, Minneapolis, Minn. Joseph T. Anderson, Ph.D., Professor, School of Public Health, University of Minnesota, Minneapolis, Minn. Francisco Grade, M.D., Professor, School of Public Health, University of Minnesota, Director, Jay Phillips Research Laboratoq, Mount Sinai Hospital, Minneap.olis, Minn.